The I gain is zero for now, but basic attitude control works.
In the video, I first change the attitude by slowly moving the joystick back and forth. At around 0:43, I move the joystick faster and let go to see how quickly the tricopter can stabilize. The P and D gains are a bit low at the moment.
I later tested untethered in the hallway (a friend took a video; I’ll include it here later), but it couldn’t correct itself fast enough to stay perfectly stable. Another problem is that the accelerometer is not mounted perfectly level with the chassis. Since I use the accelerometer to keep the DCM horizontally level, this causes the tricopter tends to tilt to one side. Finally, I currently don’t have PID control of Z-axis rotation, so I had to correct the yaw motion manually.
Notice that the attitude lags noticeably behind my joystick inputs, especially when I move the joystick quickly. The lag should be much, much less if this is going to stay completely stable in the air.